July 25, 2018 was a humid and warm day in Montreal. McGill students and postdocs in the CHIME/FRB team, us two among them, sat together in a windowless room at the university for our weekly codathon (i.e., a coding marathon), just as any other week. The one new thing: our telescope, CHIME, was collecting data that day, even though the instrument was nowhere close to design sensitivity. At around 2 pm, one of the students whispered to the person sitting next to him to ask for an opinion on a signal, just picked up by the telescope, that looked very much like a fast radio burst (FRB). The neighbour agreed that the resemblance was striking and quickly everyone in the small room got involved. We grew more excited as we were able to tick the verification checkboxes one by one and we were relieved to say “this must be it, CHIME’s first FRB!” after many years of telescope construction, software development, incessant debugging, sleepless nights and coffee.
FRBs are millisecond-duration radio transients of extragalactic origin. The first FRB was reported on in 2007, but it took until the detection of four more FRBs, in 2013, for them to become an accepted class of astronomical transients. Around this time, cosmologists in Canada were conceiving a novel radio telescope to measure the expansion history of the Universe around the time when dark energy became dominant: the Canadian Hydrogen Intensity Mapping Experiment, or CHIME. It was soon realised that CHIME’s wide field-of-view and sensitivity are excellent to search for FRBs. A second proposal, to set up this FRB search with CHIME alongside the cosmology experiment, was approved in 2015.
We started working on the CHIME/FRB project in 2016, when CHIME’s four cylindrical reflectors had just been built, but were still devoid of antennas and cabling. The goal for us students and postdocs was clear: help finish building the CHIME telescope, build a computing cluster for the FRB searching, and construct a software processing pipeline that can automatically find and characterize FRBs in real-time, from scratch. It was an opportune time to join the project with plenty of avenues to learn, contribute and lead. All of us spent multiple weeks in British Columbia’s Okanagan valley where CHIME is located. Over the two years, we laid more than a hundred kilometres of cables in the freezing cold, painted metal sheets under sky overcast by forest-fire smoke, and installed computers in the midsummer heat while avoiding sunbathing rattlesnakes.
Meanwhile, news of other FRB detections (and non-detections) kept trickling in. Most FRBs were detected at around 1.4 GHz, in the band that contains the spin-flip (hyperfine transition) of neutral hydrogen, because that is where most searches for pulsars, which are phenomenologically similar to FRBs, were undertaken. Searches at lower frequencies, 120 to 400 MHz, had drawn a blank, including one led by our colleague, Pragya Chawla. Based on the non-detections from the Green Bank Northern Celestial Cap survey, she demonstrated that FRBs are likely fainter at low radio frequencies than at higher frequencies. For a long time, we wondered if CHIME/FRB would detect any FRBs at all. Unfortunate as that might have been, the lack of FRBs at lower frequencies could lead to interesting clues about their generation and propagation — it was possible that FRBs are intrinsically not emitted at low frequencies, that perhaps their surroundings are opaque to low frequency emission, or that scattering (which is stronger at lower frequencies) would simply make the bursts undetectable to us.
A series of FRB detections that started in 2016, though, gave us more hope that CHIME/FRB would indeed detect FRBs. CHIME/FRB team member Kiyo Masui, working with other colleagues, found an FRB in data taken with the Green Bank Telescope at 800 MHz, just the top of the CHIME/FRB band (Masui et al. 2015), and this good news was followed by a handful of FRB detections from the UTMOST telescope at about 830 MHz.
The processing pipeline to find FRBs with CHIME was essentially a blank sheet when we started. CHIME/FRB had to perform real-time analysis of data at a rate hitherto unprecedented. Almost 13 terabits of data, the result of 8,000,000 measurements of the electric field every second at each of CHIME’s 1,024 antennas, has to be analyzed to find a burst that only lasts a few milliseconds. Before claiming a detection the signal from each antenna needs to be combined and focused on the sky, the data need to be calibrated, human-made radio interference (the mobile-phone network for example) and other known sources of radio emission need to be filtered out and an FRB’s characteristic sweep in frequency and time needs to be searched for.
Never prone to focus on one project, some of our team members had, outside of CHIME/FRB, detected, and after a year’s effort, accurately localized the first repeating fast radio burst source, FRB 121102. This was a huge leap for the field — the localization allowed us to find the host galaxy, measure the distance (about 1 Gpc) and understand the energetics of FRBs for the first time. One of the new lines of inquiry then became: how different is this repeater from the other, non-repeating, FRBs? What characteristics do they share? And, how common are the repeaters? It was clear that the next few repeater detections would be very valuable and that CHIME was very well suited to help answer some of these questions.
During the summer of 2018, while we finished building the telescope hardware, we detected twelve more FRBs, one of which showed repeated bursts. Even more interesting was that some of the bursts of this new source show several components that march down in frequency as time progresses, as was also observed in bursts from the first repeating FRB source. It is too soon to draw any conclusions, based on two sources, but it is an interesting parallel. The months from that first discovery leading up to the first papers have been full of verifying, checking and double-checking results and better understanding our telescope. Seeing this project through from conception to first discoveries has been extremely rewarding.
The fact that CHIME/FRB has seen FRBs down to 400 MHz, some with very little scattering, suggests that FRB surroundings are likely not very opaque at 400 MHz and some FRBs are likely detectable even at lower frequencies, based on their scattering properties. Last summer when the first detections were made, CHIME/FRB was operating at less than a quarter of its full search capacity while still detecting 13 FRBs in 3 weeks. Now, with CHIME/FRB operating at full capacity, we are preparing for a profusion of FRBs, trying to automate the processes of verification and calibration so that we can move to detailed studies of the bursts and their populations. Bit by bit (or rather terabit by terabit) and burst by burst we, as a community, can piece together the FRB puzzle.